[Purpose/significance] While AIGC strongly empowers scientific research knowledge production, it also triggers traditional academic misconduct risks such as plagiarism, fabrication and falsification, giving rise to new risk patterns such as AI ghostwriting and AI technology limited academic misconduct, which will impact on the current sci⁃entific research evaluation system. It is necessary to trace the root causes of these risks and regulate them. [Method/process] By deconstructing the underlying logic of AIGC intervention in knowledge production, it can be seen that Gen⁃erative AI acts on the knowledge level evolution path of "data-information-knowledge", as well as the thinking opera⁃tion process of knowledge search and evaluation, resulting in diverse types of Human-AI collaboration relationships,such as AI dominant, AI cooperative and AI assisted. The academic misconduct responsibilities of researchers vary un⁃der different types of Human- AI collaboration. [Result/conclusion] In terms of specific responsibility regulation
schemes, the traditional objective accountability stance is difficult to cope with the problem of "identity mismatch" be⁃tween AIGC academic misconduct subjects and responsible subjects. In this regard, the subjective accountability stance should be adopted, and set differentiated academic normative obligations such as originality declaration, trans⁃parency obligation, falsification prohibition, viewpoint tracing and factual examination for the author based on their "participation" and "contribution" in knowledge production. The author's subjective fault and its condemnability could be further evaluated by the nature and degree of obligation violations, thereby achieving a reasonable construction of the AIGC academic misconduct risk responsibility system.